v1.0

ALPR Route Planner

A map of license plate reader cameras, including Flock Safety cameras, that plans a driving route around them and compares it with the fastest route. Runs on the web and as an Android app.

Desktop app: the camera-avoiding route and the fastest route across San Luis Obispo, with mapped cameras and a comparison panel — Cal Poly to Madonna Inn: the fastest route passes 4 mapped cameras; the avoiding route passes 2 for 14 more minutes.

Phone layout: the same trip with the route comparison in a bottom sheet — Cal Poly to Madonna Inn: the fastest route passes 4 mapped cameras; the avoiding route passes 2 for 14 more minutes.

When: 2026
Stack: Node.js 22, plain JavaScript with no npm dependencies, Leaflet, OpenStreetMap (Overpass, Nominatim), Valhalla routing, Capacitor for Android
Try it: fellobello.github.io/projects/alpr-demo
Code: github.com/fellobello/alpr-route-planner

What it does

Shows ALPR cameras mapped in OpenStreetMap.
Search for a start and destination, click the map, or use your location.
Plans the fastest route and one that steers around cameras, with the camera count and extra time for each. You choose how close counts as passing a camera.

Interesting parts

Works when the data service doesn't

The public camera service (Overpass) is often slow or down. Camera data is cached by map tile and saved to disk, so areas seen before load instantly. Out-of-date tiles are still shown while a refresh runs in the background, and backup mirrors are tried in turn.

One codebase, web and Android

The API logic has no HTTP in it, so the same code runs in the Node server and inside the Android app. The phone app plans routes on the device and needs no server.

Honest about its limits

Camera locations are crowd-sourced and incomplete. A route that avoids every mapped camera can still pass unmapped ones.

How it was built

Roughcut built the first version from one line: "google maps but you can see flock cameras and plan routes around them." I then took it to a 1.0 release: the redesign, place search, offline-tolerant camera data and the Android app. The code is AI-written from my direction, with about 600 tests that never touch the network.